Reconstruction of Fire Origin and Its Strength in Fire Investigation Using Intelligent Approach

With a land mass of 1,104 km2 and a population of seven million people, Hong Kong is one of the most densely populated areas in the world. Any fire occurring in this city may incur severe consequences, including substantial property losses and fatalities. After each fire is extinguished, a fire investigator must establish the cause of the fire. In this exercise, determining the origin and strength of the fire is extremely important as it can help the fire investigator to establish whether the fire was due to an accident or arson. The identified cause of the fire may also trigger the implementation of additional preventive measures in the building or even the upgrading of statutory requirements. Traditionally, a fire investigator carries out the task according to his/her professional experience rather than a holistic scientific approach. In this study, we propose to develop an intelligent approach that will facilitate the investigation of fires. The engulfing fire and smoke spewing out from the fire compartment reveal the burning of a variety of fire sources, such as synthetic materials, that are commonly found in domestic dwellings. In most cases, these materials do not burn cleanly and thus generate large amounts of soot. The amount of soot in the smoke can be significant. Some of these soot-rich gases congregate and are deposited on the walls of the compartment, leaving a blackened pattern on the walls after the fire has been extinguished. This soot deposition pattern is unique to each fire scenario, thus acting like a fire ‘fingerprint’ that can be used to reconstruct the origin and strength of the fire. It is theoretically possible to use computational fluid dynamics (CFD) to reconstruct a similar soot deposition pattern by trying different combinations of fire strengths and locations. However, relying solely on CFD to determine the fire strength and origin is not considered practical, as each simulation requires extensive computational resources. We propose an alternative approach to identifying the fire origin and strength. First, a knowledge base will be established by collecting fire data from fire experiments, fire accident records and CFD simulations. Second, an intelligent model will be developed and trained to simulate the highly nonlinear behaviour between the soot deposition pattern and the fire origin and strength. This intelligent system will be an efficient tool to assist fire investigators in estimating the origin and strength of fires in field investigations.